Loading
Of the thousands of mutations in a tumour, only a handful drive it: a stuck accelerator (oncogene, one hit) or cut brakes (tumour suppressor, both copies). The rest are passengers, though they make the tumour visible to the immune system.
Pick a product above a diagram to see the nodes it hits and the escape routes below the block. Hover or tap any node or arrow for what it is; every node opens its target, glossary entry or the pathway page. Violet boxes are druggable targets.
A car with a stuck accelerator (oncogene: one fault is enough) and cut brake lines (tumour suppressor: both lines must fail because they are duplicated). The scratches on the paintwork (passengers) did not cause the crash but they tell you what road it drove on.
In plain words, then the glossary entries the stage rests on. Chapter 2, How a cell becomes cancer: Cancer is evolution inside a body.
Of the thousands of mutations in a tumour, only a handful drive it: a stuck accelerator (oncogene, one hit) or cut brakes (tumour suppressor, both copies). The rest are passengers, though they make the tumour visible to the immune system.
Drivers, passengers & the two-hit model. Of the thousands of mutations in a tumour, only a handful (typically 2-8) actually drive it. Drivers either jam an accelerator on (oncogenes, one hit is enough) or remove a brake (tumour suppressors, both copies must go). Everything else is a passenger along for the ride.
The proteins and genes at this stage, with their role and how many products act on each. Listed players come from the atlas; drawn players sit as nodes in the diagrams above.
A growth receptor that is mutated in some lung cancers and overproduced in others; the first great success of targeted pills.
A growth-signal receptor. Some cancers make far too much of it, and drugs that block it or use it as a docking site have transformed those cancers.
KRAS is the most commonly mutated cancer gene, called 'undruggable' for 40 years until 2021.
BRAF is a signalling kinase mutated in half of melanomas; blocking it with two drugs at once became a template for targeted therapy.
The fusion that defines chronic myeloid leukaemia and a quarter of adult acute lymphoblastic leukaemia; the first cancer driver ever switched off by a pill.
TP53 is the 'guardian of the genome', broken in half of all cancers. Fixing it directly has so far defeated every attempt, so drugs exploit what its loss makes cancers depend on.
Products grouped by the node they hit, most advanced first, with the cancers an approved product is linked to. Pick one above the diagram to see it light up.
Records tied to this stage that describe resistance, evasion or tolerance. Resistance: how tumours escape each drug class lists the routes class by class.
The first drug against the KRAS protein nearly doubled survival in pancreatic cancer in 2026, but on its own it holds the disease for months, not years. Trials are now testing it in combination with a second RAS drug and with chemotherapy, and earlier in the disease; the open questions are which combination, in which order, and what works when the tumour escapes.
If most tumours escape a drug by the same back-up route, blocking that route from the start may prevent resistance rather than chase it.
About a third of triple-negative tumours are HER2-low and so eligible for trastuzumab deruxtecan, but the difference between a HER2 score of 0 and 1+ is the one pathologists agree on least, and most triple-negative tumours were scored before the label mattered. Re-scoring archived slides with digital help when a patient relapses would find the eligible third.
Giving a targeted drug in pulses rather than continuously might slow the emergence of resistant cells and reduce side effects. Early results are mixed, so this needs careful trials with clear rules for when to try it.
Aggressive glioblastomas, sarcomas and gastric cancers keep amplified cancer genes such as EGFR, MYC, MDM2 and CDK4 on free-floating DNA circles (ecDNA) whose copy number rises and falls quickly, letting the tumour dial resistance up and down. Cells carrying ecDNA depend on CHK1, giving a first drug target.
Choose two treatments so that whatever the tumour does to escape the first, it becomes easier to kill with the second. The immune system is a good candidate partner.
The cells that survive targeted therapy change shape and become unusually dependent on an antioxidant enzyme, GPX4. Hitting them in that window might stop resistance before it evolves.
Resistance mutations often exist in a tiny fraction of cells before treatment starts. Error-corrected sequencing that detects variants below 0.01 percent allele fraction could find them at diagnosis and prompt a mechanism-matched combination from day one.
When a resistance mutation first appears in the blood, the current drug is often still controlling most of the tumour. Adding a second drug rather than swapping may keep both under control.
Use tumour DNA in the blood as the signal to pause and restart a lung cancer pill, keeping the tumour in check while slowing the rise of resistant cells.
Biomarkers, tests and assays in the corpus that read this stage in a patient.
What is not known at this stage: the atlas's own questions, the bottlenecks it bears on, and the ideas in the corpus that try to answer them.
Anyone with ovarian, pancreatic, metastatic prostate or mismatch-repair-deficient colorectal cancer should be tested for inherited mutations, yet testing rates fall well short. Making it an automatic, opt-out laboratory step triggered by pathology, as reflex mismatch-repair testing already is, would close the gap.
Pack-year rules miss people who get lung cancer without heavy smoking, including East Asian women who never smoked, as Taiwan's TALENT study showed. Eligibility by a validated risk model with a set threshold, plus a never-smoker arm where family history matters, would find more cancers per scan.
Most people carrying a high-risk cancer gene do not know it until someone in the family gets cancer. Offer testing to all adults so carriers can be protected before that happens.
One patient in twenty with pancreatic cancer carries an inherited gene fault, and most have no family history. Guidelines now say test every patient, which finds relatives who carry it too, but the yearly scans that catch cancer at stage I in carriers are still offered only in research programmes. The proposal is to make surveillance follow the test result automatically.
Sequence every cancer at diagnosis, along with the patient's inherited genes, and pool the results with treatments and outcomes so every patient teaches the system how to treat the next.
Three antibody-drug conjugates now used in triple-negative breast cancer carry the same kind of chemotherapy warhead, a topoisomerase inhibitor. Nobody has randomised which to give first or whether the second works after the first; small series suggest it often does not. With two now approved first line, the question decides what a patient gets for the rest of her life.
Companies and public funders would pool money and scientists to crack the hardest cancer proteins, such as MYC and mutant p53, sharing everything openly until there is a real drug candidate, then competing on the final product.
Drugs are processed differently by people with different genetic backgrounds, but doses are set mostly in white and East Asian populations. Every new drug should be studied for how it behaves across ancestries, with dosing advice by genotype rather than by race.
Stomach cancer now has three add-on biomarkers (HER2, PD-L1, Claudin 18.2) that often overlap. Test whether combining two add-ons beats picking one.
Add immunotherapy to the two targeted pills in the most aggressive thyroid cancer, because the combination has produced multi-year survivors in early series.
36 more ideas are linked to this stage's pathways, targets and terms; see the rankings →
Papers in the corpus tied to this stage's pathways, targets and terms, newest first.
src/data/mechanics-atlas.ts). Players, medicines, escape routes, tests, ideas and papers are resolved from the knowledge graph at build time through the stage's pathways, targets and terms, so every item here has its own page and sources. Where a section is missing, the corpus has no record tied to the stage yet. Nothing here is medical advice; see about and methodology. Stage 2.2 of 56.